51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220 | class OllamaMultiModal(MultiModalLLM):
base_url: str = Field(
default="http://localhost:11434",
description="Base url the model is hosted under.",
)
model: str = Field(description="The MultiModal Ollama model to use.")
temperature: float = Field(
default=0.75,
description="The temperature to use for sampling.",
gte=0.0,
lte=1.0,
)
context_window: int = Field(
default=DEFAULT_CONTEXT_WINDOW,
description="The maximum number of context tokens for the model.",
gt=0,
)
request_timeout: Optional[float] = Field(
description="The timeout for making http request to Ollama API server",
)
additional_kwargs: Dict[str, Any] = Field(
default_factory=dict,
description="Additional model parameters for the Ollama API.",
)
_client: Client = PrivateAttr()
def __init__(self, **kwargs: Any) -> None:
"""初始化参数和ollama客户端。"""
super().__init__(**kwargs)
self._client = Client(host=self.base_url, timeout=self.request_timeout)
@classmethod
def class_name(cls) -> str:
return "Ollama_multi_modal_llm"
@property
def metadata(self) -> MultiModalLLMMetadata:
"""LLM元数据。"""
return MultiModalLLMMetadata(
context_window=self.context_window,
num_output=DEFAULT_NUM_OUTPUTS,
model_name=self.model,
is_chat_model=True, # Ollama supports chat API for all models
)
@property
def _model_kwargs(self) -> Dict[str, Any]:
base_kwargs = {
"temperature": self.temperature,
"num_ctx": self.context_window,
}
return {
**base_kwargs,
**self.additional_kwargs,
}
def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
"""聊天。"""
ollama_messages = _messages_to_dicts(messages)
response = self._client.chat(
model=self.model, messages=ollama_messages, stream=False, **kwargs
)
return ChatResponse(
message=ChatMessage(
content=response["message"]["content"],
role=MessageRole(response["message"]["role"]),
additional_kwargs=get_additional_kwargs(response, ("message",)),
),
raw=response["message"],
additional_kwargs=get_additional_kwargs(response, ("message",)),
)
def stream_chat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponseGen:
"""流式聊天。"""
ollama_messages = _messages_to_dicts(messages)
response = self._client.chat(
model=self.model, messages=ollama_messages, stream=True, **kwargs
)
text = ""
for chunk in response:
if "done" in chunk and chunk["done"]:
break
message = chunk["message"]
delta = message.get("content")
text += delta
yield ChatResponse(
message=ChatMessage(
content=text,
role=MessageRole(message["role"]),
additional_kwargs=get_additional_kwargs(
message, ("content", "role")
),
),
delta=delta,
raw=message,
additional_kwargs=get_additional_kwargs(chunk, ("message",)),
)
def complete(
self,
prompt: str,
image_documents: Sequence[ImageDocument],
formatted: bool = False,
**kwargs: Any,
) -> CompletionResponse:
"""完成。"""
response = self._client.generate(
model=self.model,
prompt=prompt,
images=image_documents_to_base64(image_documents),
stream=False,
options=self._model_kwargs,
**kwargs,
)
return CompletionResponse(
text=response["response"],
raw=response,
additional_kwargs=get_additional_kwargs(response, ("response",)),
)
def stream_complete(
self,
prompt: str,
image_documents: Sequence[ImageDocument],
formatted: bool = False,
**kwargs: Any,
) -> CompletionResponseGen:
"""流程完成。"""
response = self._client.generate(
model=self.model,
prompt=prompt,
images=image_documents_to_base64(image_documents),
stream=True,
options=self._model_kwargs,
**kwargs,
)
text = ""
for chunk in response:
if "done" in chunk and chunk["done"]:
break
delta = chunk.get("response")
text += delta
yield CompletionResponse(
text=str(chunk["response"]),
delta=delta,
raw=chunk,
additional_kwargs=get_additional_kwargs(chunk, ("response",)),
)
async def acomplete(
self, prompt: str, image_documents: Sequence[ImageDocument], **kwargs: Any
) -> CompletionResponse:
raise NotImplementedError("Ollama does not support async completion.")
async def achat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponse:
raise NotImplementedError("Ollama does not support async chat.")
async def astream_complete(
self, prompt: str, image_documents: Sequence[ImageDocument], **kwargs: Any
) -> CompletionResponseAsyncGen:
raise NotImplementedError("Ollama does not support async streaming completion.")
async def astream_chat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponseAsyncGen:
raise NotImplementedError("Ollama does not support async streaming chat.")
|